Issue link: https://epubs.iltanet.org/i/4636
www.iltanet.org Infrastructure Technologies 7 FIRM BACKGROUND Davis LLP has seven offices across Canada and one in Tokyo. The largest office is in Vancouver, with more than 300 users, while three others have between 65 and 100 users, and the remaining offices have between five and 15 users. Each major office is interconnected using a meshed multiprotocol label switching (MPLS) network with quality of service (QoS) for the Cisco VoIP system and redundant Internet-based virtual private network (VPN) links. Small and remote offices connect solely via VPN. WAN accelerators are located in each office to optimize bandwidth and to ensure remote offices have near-LAN performance. The firm has a centralized and virtualized infrastructure with the primary datacenter in the largest office and the backup data center located in another office. VIRTUALIZATION: THEN AND NOW This article assumes the reader has general knowledge of VMware's virtualization features and functionality. A detailed explanation of virtualization along with the evaluation, transition and implementation of the virtualized infrastructure and initial DR plan at Davis, is covered in an article titled "Cost-Effective Virtualization from Desktop to Data Center" in ILTA's June 2007 white paper on virtualization. Since the original implementation, the following changes have occurred: • Unclustering Microsoft Servers Due to poor failover and failback performance and with the resiliency of VMware ESX with VMotion, we reduced the complexity of the core servers by unclustering them. Clustered 2003 vintage servers require a shared read and write (RAW), not virtualized, quorum and data disks. Because of this, virtualization tools such as snapshot and replication cannot be used in a clustered setup. • WAN Accelerators Cisco Wide Area Application Service (WAAS) accelerators were installed to improve bandwidth utilization, optimize traffic and compress data, all of which improved the user experience while reducing WAN costs. • Virtual Machine (VM) Creep Since the initial implementation of the nine virtualized servers, we now have 50 to 60 development and production VMs in a fully virtualized environment. • EVault Replication We enhanced disk-to-disk backup resiliency by adding replication of backup data to the DR site using EVault's proprietary replication technology. ORIGINAL HARDWARE The original configuration was a pair of IBM quad processor servers, each running VMware ESX connected to an IBM-branded NetApp SAN via Fibre Channel. Secondary servers were connected via iSCSI to reduce the cost of the Fibre Channel infrastructure. Shared storage was used, as it is a requirement for VMware VMotion, which facilitates redundancy in the event of a hardware failure or maintenance window. Located at the DR site was a less powerful server also running ESX with a direct attached storage (DAS) shelf. Virtualization is hardware